[ https://issues.apache.org/jira/browse/ARROW-3613?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16680379#comment-16680379 ]
Alexandre Crayssac commented on ARROW-3613: ------------------------------------------- Hello, After digging into the code I confirm that the field `length` of `Float64Builder.builder` is not updated when calls to `Float64Builder.Resize` are made. Note that it is also the case with all the numeric types since they are all generated by a common template and they all use the `builder` field to retrieve the length. By the way, should not we use the field `data` to retrieve the lenght? It looks like the problem is rooted in the function `resize` of the `builder` field which does not update its `length` field. Note that the `capacity` field is updated so I think we might end up with `length > capacity`. (Note that the length of `data` is correctly updated.) Can someone confirm that's a bug? Disclaimer: I don't have tons of experience with Go and Arrow. Alexandre > [Go] Resize does not correctly update the length > ------------------------------------------------ > > Key: ARROW-3613 > URL: https://issues.apache.org/jira/browse/ARROW-3613 > Project: Apache Arrow > Issue Type: Bug > Components: Go > Reporter: Jonathan A Sternberg > Priority: Major > > If you have the following code: > {code:java} > package main > import ( > "fmt" > "github.com/apache/arrow/go/arrow/array" > "github.com/apache/arrow/go/arrow/memory" > ) > func main() { > builder := array.NewFloat64Builder(memory.DefaultAllocator) > fmt.Println(builder.Len(), builder.Cap()) > builder.Reserve(44) > fmt.Println(builder.Len(), builder.Cap()) > builder.Resize(5) > fmt.Println(builder.Len(), builder.Cap()) > builder.Reserve(44) > for i := 0; i < 44; i++ { > builder.Append(0) > } > fmt.Println(builder.Len(), builder.Cap()) > builder.Resize(5) > fmt.Println(builder.Len(), builder.Cap()) > } > {code} > It gives the following output: > {code:java} > 0 0 > 0 64 > 0 32 > 44 64 > 44 32 > {code} > For whatever reason, the length is not recorded as 5. I understand why the > capacity might not be 5, but it does seem like the length should be set to 5 > if the array is resized to a length smaller than its current capacity. -- This message was sent by Atlassian JIRA (v7.6.3#76005)